https://izabellaleahz.github.io/biopapers/New papers from watched PIs, and papers getting attentionhttps://arxiv.org/abs/2610.2227610.48550/arxiv.2610.22276arXiv — Spatial intelligence lets an agent perceive and act in the physical world, and many of its judgments must be made at interaction rates: which spatial relation holds, how far an object is, or where to place it. Vision-language models (VLMs) produce these answers by free-form text generation, paying for autoregressive decoding and answer parsing even though the output space is compact: one of a few options, a scalar, or a 2D point. In light of this, we introduce MJev-Spatial, a System 1 multimodalTue, 20 Oct 2026 00:00:00 GMThttps://arxiv.org/abs/2610.2220010.48550/arxiv.2610.22200arXiv — Computer use has reached a turning point: the focus is expanding from GUI-only interaction to systems that complete tasks on users' computers through visual and programmatic interfaces. We define a generalist computer-use agent (GCUA) as a system in which the model chooses how to observe, act, and organize its work across visual and programmatic interfaces. Yet frontier implementations, such as GPT-6 Astra's computer-use system, remain largely closed, limiting systematic study of their design. WTue, 20 Oct 2026 00:00:00 GMThttps://www.science.org/doi/abs/10.1126/science.adt1928?af=R10.1126/science.adt1928Science — Intestinal homeostasis requires tight regulation of stem cell maintenance and differentiation to absorptive and secretory cell types, two major intestinal lineages. Key signaling pathways have been identified, but how they balance cell type composition remains unclear. We uncovered oscillatory expression of the Notch-target protein Hes1 (hairy and enhancer of split 1) that controls cell fate specification in a frequency-dependent manner. An endogenous fluorescent Hes1 reporter revealed distinct,Tue, 20 Oct 2026 00:00:00 GMThttps://www.science.org/doi/abs/10.1126/science.aee6550?af=R10.1126/science.aee6550Science — Fuel economy matters to both animals and machines, as the cost of transport can determine where, when, and how far locomotion is possible. However, established estimates of these values for animal flight come from laboratories and treat the atmosphere as constant, even though free-flying animals routinely extract energy from atmospheric flows. Using high-resolution movement data, we estimate the cost of transport for 50 bird and bat species flying in the wild. Our results revise the classic rankTue, 20 Oct 2026 00:00:00 GMThttps://www.nature.com/articles/d41586-026-03266-x10.1038/d41586-026-03266-xNature — Nature, Published online: 14 October 2026; doi:10.1038/d41586-026-03266-x A coral fragment can form a tissue bridge with a dying counterpart — and provide it with crucial microalgae.Tue, 20 Oct 2026 00:00:00 GMThttps://arxiv.org/abs/2610.2057110.48550/arxiv.2610.20571arXiv — Long-form agentic video generation must turn a screenplay into shots that remain spatially and temporally consistent across nonconsecutive requests. Passing prompts or recent frames between shots does not provide a shared spatial frame for revisiting a scene or a time-indexed record for replaying an earlier event. We present Any Time, Any Space (ATAS), which compiles a screenplay into shot requests with explicit location, camera, and event-time references. Before generation, it retrieves relevanTue, 20 Oct 2026 00:00:00 GMThttps://arxiv.org/abs/2610.2058410.48550/arxiv.2610.20584arXiv — Computer-use agents have demonstrated strong capabilities in interpreting screens and executing keyboard and mouse actions. However, games, with their visually rich worlds and diverse mechanics, remain challenging and offer a demanding next step toward general digital control. Current game-agent learning pipelines often rely on large-scale demonstrations with limited coverage of dynamic game states, while scalable environments for interactive training remain difficult to obtain. We introduce PlaTue, 20 Oct 2026 00:00:00 GMThttps://doi.org/10.1038/s41591-026-04717-810.1038/s41591-026-04717-8Nature Medicine — Tue, 20 Oct 2026 00:00:00 GMThttps://arxiv.org/abs/2610.1886110.48550/arxiv.2610.18861arXiv — How can we build an agent that thinks beyond human language, yet still acts through it in the external environment? Modern LLM agents still reason and interact through text, while latent reasoning offers an efficient internal medium but lacks native tool access and reasoning transparency. We bridge these two directions with Hybrid Reasoning Agents (HybridAgents), which dynamically navigate between latent reasoning and text-based interaction throughout a unified generation process. ArchitecturallTue, 20 Oct 2026 00:00:00 GMThttps://arxiv.org/abs/2610.1877310.48550/arxiv.2610.18773arXiv — System one models enable direct, low-latency decision-making. Yet recent practice of predicting an answer from serialized text does not, by itself, respect the structure of a decision problem. We argue that current designs underemphasize four important structural requirements. (i) null-aware: Models should distinguish answer absence from uncertainty among candidates; (ii) permutation-invariant: Reordering alternatives should not change the decision; (iii) option-scalable: Growing candidate sets Tue, 20 Oct 2026 00:00:00 GMThttps://arxiv.org/abs/2610.1861510.48550/arxiv.2610.18615arXiv — Visual agents must turn video observations into timely decisions to track progress and choose what to do next. Building on the emerging System One paradigm, we introduce VideoJev, a family of models that cast video understanding and interaction as structured decision tasks. We adapt pretrained multimodal large language models (MLLMs) by replacing autoregressive answer decoding with a shared head that scores temporal intervals and textual options against video evidence. These scores yield event sTue, 20 Oct 2026 00:00:00 GMThttps://doi.org/10.1038/s41598-026-74792-x10.1038/s41598-026-74792-xScientific Reports — Peripheral blood mononuclear cells (PBMCs) profiled in different data modalities have emerged as diagnostic and prognostic biomarkers for cancer. Here we demonstrate that PBMC chromatin images contain sufficient information to track patients’ trajectories during and after radiation therapy. We collected images from 150 patients across a variety of cancer types including Central Nervous System (CNS) cancer, Head & Neck cancer, Chordoma and Chondrosarcoma, Sarcoma, and Lymphoma, and adapted a multTue, 20 Oct 2026 00:00:00 GMThttps://arxiv.org/abs/2610.1257210.48550/arxiv.2610.12572arXiv — Online reinforcement learning (RL) fine-tunes diffusion models with black-box rewards, but reverse-process methods such as Flow-GRPO are limited in efficiency by poor credit assignment, while forward-process methods such as DiffusionNFT are efficient yet unstable without carefully tuned exponential-moving-average (EMA) off-policy training. We show that NFT-style objectives contain a hidden forward--reverse consistency force that joint positive--negative training cancels only partially, and that Tue, 20 Oct 2026 00:00:00 GMThttps://doi.org/10.64898/2026.10.06.75688910.64898/2026.10.06.756889bioRxiv — Folate is an essential driver of one-carbon metabolism. The cellular consequences of its deprivation in CD8 + T cell and impact on anti-tumor function are unknown. Single cell RNA-seq analysis of primary central nervous system lymphoma (PCNSL) patients’ tumor samples revealed exhaustion in T cells with high folate metabolism gene expression signature. Functional genomics, metabolite profiling, and glycomics identified disrupted cell surface glycosylation as a major consequence of folate deprivatTue, 20 Oct 2026 00:00:00 GMThttps://doi.org/10.1093/pnasnexus/pgag30610.1093/pnasnexus/pgag306PNAS Nexus — Women faculty experience academia differently from men in many ways, which can lead them to consider leaving their positions. Using a large-scale survey of both current and former tenure-line faculty across nearly all US PhD-granting institutions and 29 disciplines, we investigate which aspects of the academic workplace are (i) most gendered and (ii) most relevant to faculty retention patterns among respondents. Analyzing 6,615 free-text responses reveals that feeling devalued at work, both in fTue, 20 Oct 2026 00:00:00 GMThttps://arxiv.org/abs/2609.3910210.48550/arxiv.2609.39102arXiv — Self-evolving search agents build their own training curricula by jointly optimizing a proposer that generates questions and a solver that answers them. This closed loop introduces a failure mode we call co-cheating: the proposer and solver increasingly agree on shared errors, so internal reward improves without a matching gain in external correctness. A post-hoc audit against source evidence shows co-cheating growing more severe over successive rounds of self-evolution, with pseudo-label correcTue, 20 Oct 2026 00:00:00 GMThttps://arxiv.org/abs/2609.3960110.48550/arxiv.2609.39601arXiv — Precise grounding matters. It specifies which object is the target and where that object is, even in clutter and for tiny objects, and it has to be fast enough for closed-loop control. Yet vision-language-action (VLA) and world-action models (WAMs) take perception from general-purpose vision-language and video-generation backbones, which still fail in these settings. We introduce GroundingPI, a 4B grounding foundation model that generates points and boxes as quantized coordinates in a shared vocTue, 20 Oct 2026 00:00:00 GMThttps://arxiv.org/abs/2609.3648410.48550/arxiv.2609.36484arXiv — On-policy distillation (OPD) trains a student to match the teacher's next-token distributions on the student's own trajectories and has yielded substantial empirical gains. Generalized variants allow the student to surpass the teacher by extrapolating an implicit reward in output space. The language-model head, however, attenuates this change anisotropically: much of the change encoded in the teacher's hidden states reaches the logits at a small fraction of its weight, and the sampled-token log-Tue, 20 Oct 2026 00:00:00 GMThttps://arxiv.org/abs/2609.3842610.48550/arxiv.2609.38426arXiv — We introduce LoopVL to study whether Loop Transformers can be effectively extended to vision- language models. LoopVL combines Module-Loop and Model-Loop computation to iteratively update a unified vision-language state through shared modules. We train LoopVL from scratch through language pre-training, multimodal training, and post-training. LoopVL outperforms a range of similarly sized and larger non-recurrent models on multimodal understanding and visual reasoning benchmarks. We also observe VTue, 20 Oct 2026 00:00:00 GMThttps://arxiv.org/abs/2609.3430910.48550/arxiv.2609.34309arXiv — Executable programs offer explicit control over how images and videos are constructed, but generating runnable code is only the beginning of visual creation. A program can execute correctly while violating the requested composition, appearance, or motion. We define this discrepancy as the Program-to-Visual (P2V) gap and introduce MaLiang-Harness, a unified framework for organizing MLLM-driven visual generation into a persistent process of construction, inspection, and revision. Its central desigTue, 20 Oct 2026 00:00:00 GMThttps://arxiv.org/abs/2609.3601210.48550/arxiv.2609.36012arXiv — General-purpose robots must infer what a new task requires and translate that understanding into appropriate physical action. In-context learning (ICL) for robots supports this process by using demonstrations and interaction to direct existing competence with neural parameters held fixed during deployment. We organize this literature review around the interfaces connecting contextual evidence to execution, distinguishing four families: context-conditioned policies, geometric demonstration transfTue, 20 Oct 2026 00:00:00 GMThttps://arxiv.org/abs/2609.3343910.48550/arxiv.2609.33439arXiv — As large language models advance, AI agents are moving beyond isolated, domain-specific tasks toward long-horizon, cross-domain workflows. This transition exposes two challenges: increasing harness complexity makes manual design difficult to scale, while tighter coupling to specific domains limits the generality of a single harness. The central question thus shifts from how to engineer a stronger harness for one domain to how to autonomously construct specialized harnesses, improve them through Tue, 20 Oct 2026 00:00:00 GMThttps://arxiv.org/abs/2609.3272210.48550/arxiv.2609.32722arXiv — *Reinforcement learning (RL)* can induce substantial reasoning capabilities in large language models (LLMs), but how much of this capability transfers across model scales, and how quickly, remains unclear. We study the scaling properties of *on-policy distillation (OPD)* across *weak-to-strong*, *same-base*, and *strong-to-weak* teacher--student setups. We find that early OPD training dynamics uniformly exhibit a regular *useful-transfer* regime, in which held-out accuracy (the *gold score*, G) Tue, 20 Oct 2026 00:00:00 GMThttps://arxiv.org/abs/2609.3184710.48550/arxiv.2609.31847arXiv — General-purpose agents can plan, reason, and act over long horizons, yet their production capabilities remain fragmented across text, images, audio, video, documents, 3D assets, and code. Extending a foundation model to additional modalities ties capability growth to costly model updates, while assembling specialist models and tools leaves unresolved how procedures, dependencies, intermediate assets, and cross-turn revisions should be coordinated. We present Omni-IO Skills, a plug-and-play AgentTue, 20 Oct 2026 00:00:00 GMThttps://doi.org/10.1038/s41467-026-78063-110.1038/s41467-026-78063-1Nature Communications — The interaction between T cell receptors (TCRs), peptides, and human leukocyte antigens (HLAs) underlies antigen-specific T cell immunity. Despite substantial advances in peptide– HLA presentation prediction, accurate modeling of coupled TCR–peptide–HLA recognition remains underdeveloped, limiting applications such as TCR and neoepitope prioritization in cancer and antigen identification in autoimmunity. Here we present StriMap, a unified framework for predicting TCR–peptide–HLA interactions by Tue, 20 Oct 2026 00:00:00 GMThttps://doi.org/10.1126/sciadv.adz690710.1126/sciadv.adz6907Science Advances — The RNF43 p.G659fs ( RNF43 659mut ) mutation occurs in 8% of colorectal cancers (CRCs) and is enriched in microsatellite instability–high (MSI-H) tumors. Although RNF43 659mut promotes tumor growth through phosphatidylinositol 3-kinase/protein kinase B/mechanistic target of rapamycin (PI3K/AKT/mTOR) activation independent of WNT signaling, its immunologic effects are poorly defined. We show that RNF43 659mut shapes an immunosuppressive natural killer (NK) cell landscape in MSI-H CRC. Single-cellTue, 20 Oct 2026 00:00:00 GMThttps://doi.org/10.64898/2026.09.18.75274110.64898/2026.09.18.752741bioRxiv — Glioblastomas (GBMs) are uniformly fatal brain tumors resistant to immunotherapy, yet the mechanisms driving immune dysfunction remain poorly understood. Here we identify a tissue-specific mechanism of dendritic cell (DC) dysfunction contributing to GBM immune evasion. We demonstrate that while systemic immunity depends on type 1 conventional DCs (cDC1s), GBMs preferentially accumulate cDC2s with impaired antigen-presenting capacity. Further, microglia-derived GAS6 and PROS1 signaling through thTue, 20 Oct 2026 00:00:00 GMThttps://doi.org/10.1016/j.euroneuro.2026.11302510.1016/j.euroneuro.2026.113025European Neuropsychopharmacology — Gene transcription is a highly orchestrated process driven by the interplay between genetic variation and epigenetic modifications. Transcriptomic imputation has advanced the biological fine mapping of GWAS of complex traits; models of genetically-regulated expression (GReX) are used to conduct the transcriptome-wide association studies (TWAS) and detect gene-trait-associations (GTAs) with tissue- and cell-type resolution. This static approach inherently fails to also account for the dynamic epiTue, 20 Oct 2026 00:00:00 GMThttps://doi.org/10.64898/2026.09.14.75153010.64898/2026.09.14.751530bioRxiv — Lung adenocarcinoma, driven frequently by KRAS and p53 mutations, remains a leading cause of cancer mortality. Current state-of-the-art genetically engineered mouse models often rely on viral delivery of recombinases, such as Cre recombinase, to initiate transformation. However, viral particles can infect and activate innate immune cells, thus potentially impacting studies of tumor-immune dynamics. Here, we develop a layer-by-layer (LbL) polyplex platform using poly({beta}-aminoester) (PBAE) polTue, 20 Oct 2026 00:00:00 GMThttps://doi.org/10.64898/2026.09.13.75128010.64898/2026.09.13.751280bioRxiv — Introduction Class II lupus nephritis (LN) is considered clinically benign, yet up to 50% of patients progress to more severe disease and no molecular characterization exists. We aimed to define the single-cell landscape of Class II LN and identify cellular and transcriptional features associated with kidney outcome. Methods We performed single-cell RNA sequencing on kidney biopsies from 15 Class II LN patients (8 de novo, 7 regressed from prior proliferative or membranous disease) and 6 healthyTue, 20 Oct 2026 00:00:00 GMThttps://doi.org/10.1084/jem.2026137010.1084/jem.20261370The Journal of Experimental Medicine — Scientific independence alone does not create a sustainable research program. Drawing on the 2026 Irving Cancer Immunology Symposium, we offer practical strategies for navigating uncertainty and limited resources to build a coherent research program and resilient team across funding systems.Tue, 20 Oct 2026 00:00:00 GMThttps://doi.org/10.1016/j.medj.2026.10126610.1016/j.medj.2026.101266Med — Tue, 20 Oct 2026 00:00:00 GMThttps://doi.org/10.1126/scitranslmed.aec051210.1126/scitranslmed.aec0512Science Translational Medicine — Lupus nephritis (LN), a severe manifestation of systemic lupus erythematosus (SLE), is a heterogeneous disease driven by diverse immune and tissue cell types. We obtained 538,194 single-cell and 142,881 single-nuclear profiles from kidney biopsies of 155 patients with LN and 30 preimplantation transplant biopsy controls, along with 327,326 single-cell blood profiles. We characterized key stromal and immune cell types and cell states; moreover, we distinguished cell states that were tissue specifTue, 20 Oct 2026 00:00:00 GMThttps://doi.org/10.1016/j.immuni.2026.07.01810.1016/j.immuni.2026.07.018Immunity — Tue, 20 Oct 2026 00:00:00 GMThttps://arxiv.org/abs/2608.1085710.48550/arxiv.2608.10857arXiv — Determining the complexity, or Intrinsic Dimension (ID), of data is fundamental to efficient and interpretable representation learning. This is particularly challenging in multi-modal settings when trying to learn disentangled representations for shared and private information. Existing techniques leave a critical gap: they are often static, uni-modal, or in the case of contrastive methods, adapt only to the shared ID implicitly. We introduce Fidelity-Guided Rank Optimization (FiGuRO), a framewoTue, 20 Oct 2026 00:00:00 GMThttps://doi.org/10.64898/2026.08.06.2635942210.64898/2026.08.06.26359422medRxiv — Microsatellite stable (MSS) pancreatic ductal adenocarcinoma (PDAC) is refractory to immune checkpoint blockade. We conducted a single-arm phase II trial ( NCT04361162 ) combining nivolumab, ipilimumab and radiation therapy to treat patients with pre-treated metastatic MSS PDAC (n=30). We integrated longitudinal profiling of 32 pre- and on-treatment tumor biopsies from 22 patients, yielding 245,529 single-nucleus and 128,295 single-cell transcriptomes including 27,215 T-cells with paired TCR cloTue, 20 Oct 2026 00:00:00 GMThttps://doi.org/10.1093/biosci/biag12510.1093/biosci/biag125BioScience — Earth’s climate system is moving deeper into uncharted territory. We present the 2026 update of planetary vital signs that track the state of Earth’s climate system, alongside indicators of the human activities driving ongoing change. Of 16 climate-related indicators, seven are currently at record extremes. Several indicators also show evidence of accelerating trends or abrupt shifts, raising concerns about increasing climate instability, interacting feedbacks, and the growing risk of crossing cTue, 20 Oct 2026 00:00:00 GMThttps://doi.org/10.64898/2026.07.31.74126510.64898/2026.07.31.741265bioRxiv — Paired histopathology and spatial transcriptomics data are advancing our understanding of tissue biology and disease, but modeling both modalities at single-cell resolution while mapping local and distal cell-cell interdependencies remains computationally prohibitive. Here we introduce TissueFormer, a framework for pretraining foundation models with linear rather than quadratic computational complexity, overcoming a long-standing barrier to modeling long-range dependencies at scale. Trained on oTue, 20 Oct 2026 00:00:00 GMThttps://doi.org/10.20944/preprints202608.0213.v110.20944/preprints202608.0213.v1Preprints.org — Scientific discovery is fundamentally a problem-solving process involving distributed intelligence. Human intuition, computational reasoning, and experimental execution are distributed across people, instruments, and software systems, limiting the speed and scale of discovery. Although automation, high-throughput experimentation, foundation models, and cloud infrastructure have accelerated individual stages of the scientific workflow, they have not unified the discovery process. We hypothesize tTue, 20 Oct 2026 00:00:00 GMThttps://doi.org/10.1016/j.ard.2026.07.01410.1016/j.ard.2026.07.014Annals of the Rheumatic Diseases — OBJECTIVES: Lupus nephritis (LN) is a severe complication of systemic lupus erythematosus (SLE), leading to progressive renal fibrosis and functional decline. Understanding the interplay between immune cells and stromal cells is needed to develop effective therapeutic strategies. In this study, we investigated the landscape of macrophage-fibroblast interactions in human LN and validated these findings in mouse models. METHODS: We characterised distinct fibroblast subsets and their interactions wTue, 20 Oct 2026 00:00:00 GMThttps://doi.org/10.3150/26-bej197010.3150/26-bej1970Bernoulli — We study the problem of maximum likelihood estimation given one sample (n=1) over Brownian Motion Tree Models (BMTMs), a class of Gaussian models on trees. BMTMs are often used as a null model in phylogenetics, where the one-sample regime is common. Specifically, we show that, for any fixed tree, almost surely, the one-sample BMTM maximum likelihood estimator (MLE) exists, is unique, and corresponds to a fully observed tree. Moreover, we provide a polynomial time algorithm for its exact computatTue, 20 Oct 2026 00:00:00 GMThttps://doi.org/10.1016/j.cell.2026.07.01610.1016/j.cell.2026.07.016Cell — Immune aging is being increasingly recognized as a critical barrier to effective cancer immunotherapy, as the aged tumor microenvironment (TME) drives T cell dysfunction and impairs immune control of cancer 1–4 . However, the key molecular drivers of this process as well as potential targets to rescue T cell dysfunction in aged tumors remain incompletely understood. Therefore, we performed in vivo single-cell CRISPR screens in CD8⁺ T cells within aged tumors and tumor-draining lymph nodes (tdLNsTue, 20 Oct 2026 00:00:00 GMThttps://doi.org/10.1093/jimmun/vkag141.53310.1093/jimmun/vkag141.533The Journal of Immunology — Introduction The alpha-variant wave of the COVID-19 pandemic provided a unique opportunity to study, at single-cell resolution, how near-universal exposure to the same pathogen can lead to either effective or dysfunctional immune responses in humans. Methods We analyzed 2.5 million circulating immune cells from 428 patients across time points (840 PBMC samples), encompassing three contemporaneous SARS-CoV-2 cohorts: acutely infected patients at five WHO disease severity levels and three time poiTue, 20 Oct 2026 00:00:00 GMThttps://doi.org/10.1093/jimmun/vkag141.151710.1093/jimmun/vkag141.1517The Journal of Immunology — Introduction Tumor neoantigens have emerged as promising candidates for personalized immunotherapies, however, confirming their presentation and immunogenicity remains challenging due to low neoantigen expression and limited detection sensitivity. Moreover, the spatial organization of neoantigen-specific T cell clones relative to their target tumor cells and other cell types within the tumor microenvironment remains incompletely defined due to technological limitations. To bridge this gap, we deTue, 20 Oct 2026 00:00:00 GMThttps://doi.org/10.1093/jimmun/vkag141.167410.1093/jimmun/vkag141.1674The Journal of Immunology — Introduction Human leukocyte antigen (HLA) class I presents intracellular peptides to the immune system on the cell surface. Since this process is crucial for the recognition of cancer cells and the initiation of anti-tumor immunity, peptides presented by HLA are valuable immunotherapy targets. More stable peptide HLA (pHLA) complexes provoke superior immune responses. However, how peptide sequence motifs contribute to pHLA stability is not well understood. Methods We developed a high-throughputTue, 20 Oct 2026 00:00:00 GMThttps://doi.org/10.1093/jimmun/vkag141.81310.1093/jimmun/vkag141.813The Journal of Immunology — Introduction Autoimmunity is defined by disruption of the immune system, where instead of defending the body, the immune system turns on itself and attacks its own organs instead. While this immune dysfunction connects autoimmune diseases, each disease is distinguished by the organs involved and clinical features that characterize its presentation. The goals of this study are to utilize single-cell sequencing technologies to characterize peripheral immune cells and identify shared and unique mecTue, 20 Oct 2026 00:00:00 GMThttps://doi.org/10.64898/2026.07.27.73936410.64898/2026.07.27.739364bioRxiv — Specialized cell types layer cell-type-restricted proteins, metabolites, and organelles onto a shared foundation of core cellular processes. How this ubiquitous machinery generates lineage-specific outputs is central to both cell biology and therapeutic development. G-protein-coupled receptors (GPCRs) respond to receptor-restricted ligands to drive cell-type-specific transcription through cAMP-mediated activation of protein kinase A (PKA), which activates the transcription factor CREB1 through tTue, 20 Oct 2026 00:00:00 GMThttps://doi.org/10.1093/jimmun/vkag141.00910.1093/jimmun/vkag141.009The Journal of Immunology — Introduction Resident memory T cells (TRM) mediate localized immunity in barrier tissues while central memory T cells (TCM) recirculate through lymphoid organs to surveil for reinfection. Although TRM are classically associated with peripheral non-lymphid tissues, they have also been identified within lymph nodes (LNRM) where mechanisms guiding their formation and their role during secondary pathogen encounters remain poorly understood. Moreover, functional distinctions between TCM, LNRM, and luTue, 20 Oct 2026 00:00:00 GMThttps://doi.org/10.1038/s41587-026-03194-110.1038/s41587-026-03194-1Nature Biotechnology — Tue, 20 Oct 2026 00:00:00 GMThttps://doi.org/10.64898/2026.07.15.73880910.64898/2026.07.15.738809bioRxiv — Single-cell RNA sequencing has transformed our understanding of cellular systems, yet the reliance on short-read sequencing restricts analysis to gene-level quantification and obscures the immense biological diversity generated by alternative splicing. While long-read sequencing technologies can capture full-length RNA and resolve transcript isoforms, current platforms remain constrained by throughput and high per-base costs, rendering them impractical for modern million-cell applications. To adTue, 20 Oct 2026 00:00:00 GMThttps://arxiv.org/abs/2607.1678910.48550/arxiv.2607.16789arXiv — Real-world perception and decision making are inherently multimodal, integrating complementary signals across modalities. However, training multimodal models faces two main obstacles. First, collecting large-scale, well-aligned paired multimodal datasets is often impractical, making end-to-end multimodal training difficult. Second, existing multimodal representations frequently entangle information shared across modalities with modality-specific information, hindering interpretability and controTue, 20 Oct 2026 00:00:00 GMThttps://arxiv.org/abs/2607.1181610.48550/arxiv.2607.11816arXiv — Causal discovery algorithms learn a network that describes the causal dependencies among random variables. A common workflow involves first utilizing conditional independence properties on observational data to determine partially directed causal relationships, then applying interventions to orient the unknown causal directions. A critical assumption for the first step is faithfulness: a requirement that causally linked variables exhibit statistical dependence. Many natural systems include buffeTue, 20 Oct 2026 00:00:00 GMThttps://doi.org/10.1016/j.xcrm.2026.10291010.1016/j.xcrm.2026.102910Cell Reports Medicine — The addition of aPD1 to 5-FU/platinum in advanced gastric cancer (GC) yields variable responses. To understand cooperativity between chemotherapy and immunotherapy, we previously reported a phase II trial sequentially adding pembrolizumab to 5-FU/platinum. In this study, we use single-cell RNA- and TCR-sequencing to analyze 66,813 T cells from primary tumor biopsies pre-treatment, post-chemotherapy, and post-immunotherapy in 33 patients. We observed greater abundance, persistence, and recruitmenTue, 20 Oct 2026 00:00:00 GMThttps://doi.org/10.64898/2026.06.27.73278410.64898/2026.06.27.732784bioRxiv — Chronic hepatitis B is characterized by a decades-long evolving engagement between host immunity and the hepatitis B virus (HBV). Understanding the molecular characteristics of HBV-specific CD8 T cells linked to control of viral replication and antigenemia is essential to design effective immunotherapeutic modalities. Here we show that HBV-specific CD8 T cells, even during infection stages with extremely high viral loads, lack the features of terminally exhausted CD8 T cells observed in chronic Tue, 20 Oct 2026 00:00:00 GMThttps://arxiv.org/abs/2604.0377910.5281/zenodo.20882785arXiv — Datasets, model checkpoints, precomputed missingness masks, and camera-ready evaluation results (results_*.json) to reproduce the results in CountsDiff (ICML 2026). Extract into the repository's data/ directory via scripts/download_data.sh. Code: https://github.com/rsoatto/countsdiff · Paper: arXiv:2604.03779Tue, 20 Oct 2026 00:00:00 GMThttps://doi.org/10.1126/scitranslmed.aed308710.1126/scitranslmed.aed3087Science Translational Medicine — In chronic inflammatory diseases, maladaptive tissue remodelling is driven by a complex interplay of resident cells, immune filtrates and the extracellular matrix. In the autoimmune disorder rheumatoid arthritis (RA), synovial tissue undergoes assive expansion to form an invasive pannus that drives the erosion of cartilage and bone. The mechanisms mediating this ggressive growth are incompletely defined. Using spatial transcriptomics profiling of patient tissue, we detected an bundance of prolifTue, 20 Oct 2026 00:00:00 GMThttps://doi.org/10.1038/s41467-026-74790-710.1038/s41467-026-74790-7Nature Communications — Many cancer patients treated with immune checkpoint blockade (ICB) do not have durable treatment responses. Circulating biomarkers have the potential to identify patients with primary resistance or early progression on therapy to alter treatment course and potentially avoid unnecessary toxicity. Unbiased multimodal proteomic profiling in blood has been underexplored due to the previously limited scalability of multiplexing technologies or cohorts lacking time-series sampling. To address this, weTue, 20 Oct 2026 00:00:00 GMThttps://doi.org/10.64898/2026.06.14.73216210.64898/2026.06.14.732162bioRxiv — Stem-like TCF7+ CD8 T cells sustain anti-tumor responses and support immune checkpoint blockade. We systematically identified regulators of this cell state using genome-wide CRISPR screens in primary T cells in vitro . Using random barcodes to link clonal relationships with guide identity and transcriptional states in single cells, we inferred differentiation trajectories and differentiation rates of CD8 T cells in tumors, while mitigating confounding clonal bias. We found that Trim28-deficient Tue, 20 Oct 2026 00:00:00 GMThttps://doi.org/10.1016/j.ccell.2026.05.01310.1016/j.ccell.2026.05.013Cancer Cell — Tue, 20 Oct 2026 00:00:00 GMThttps://doi.org/10.64898/2026.05.29.72855310.64898/2026.05.29.728553bioRxiv — CD4+ T cells are orchestrators of the immune system with diverse effector functions. Their full molecular diversity remains unclear due to the lack of a unified framework. Within the immgenT project, we profiled RNA, surface markers, and TCR clonotypes in conventional CD4+ T cells across >700 samples. Integration with a joint RNA-protein deep generative model revealed an ensemble of 20 CD4 states that account for all cells across tissues and challenges. Small, highly polarized clusters that evokTue, 20 Oct 2026 00:00:00 GMThttps://doi.org/10.1016/j.celrep.2026.11745510.1016/j.celrep.2026.117455Cell Reports — Melanoma is a common and aggressive cancer, with rising incidence in most developed countries. Major discoveries in melanoma biology have been rapidly translated, allowing cures for patients in late-stage disease. Despite these advances, many tumors remain refractory, in part due to an incomplete understanding of the genes and pathways gained or lost during melanoma tumorigenesis. To address this gap and provide a broadly useful resource for the scientific community, we established melPDomiX, a Tue, 20 Oct 2026 00:00:00 GMThttps://doi.org/10.64898/2026.05.21.72686310.64898/2026.05.21.726863bioRxiv — Perturbations of genes with functional importance in T cells could be used to change the distribution of CD8 T cell states to enhance anti-tumor functions for cancer immunotherapies. We launched a world-wide computational challenge to predict the effects of gene perturbations and to devise objective functions for prioritizing gene perturbations that lead to desired T-cell state distributions. We supported the challenge by generating a single-cell Perturb-seq dataset profiling the effect of knockTue, 20 Oct 2026 00:00:00 GMThttps://doi.org/10.1093/ajrccm/aamag162.617110.1093/ajrccm/aamag162.6171American Journal of Respiratory and Critical Care Medicine — Rationale Immunomodulatory agents have been applied to sepsis with mixed results but were beneficial in hypoxemic COVID-19 patients; hypoxemia may be a surrogate for immune dysregulation, but more precise predictors are needed. The sepsis-response signature (SRS) leveraged bulk RNA sequencing to identify two reproducible sepsis endotypes associated with prognosis and a differential response to corticosteroids. We hypothesized the SRS-1 endotype may identify patients with treatable immune traits.Tue, 20 Oct 2026 00:00:00 GMThttps://doi.org/10.1016/j.ard.2026.03.00410.1016/j.ard.2026.03.004Annals of the Rheumatic Diseases — Objectives Lupus nephritis (LN) is a common, potentially fatal manifestation of systemic lupus erythematosus. We aim to gain new insights into the immune responses underlying LN and their relation to the histologic heterogeneity observed in this disease, focusing on myeloid cells. Methods Single-cell RNA-sequencing (scRNA-seq) was used to profile dissociated kidney samples from 156 LN patients and 30 healthy individuals. Spatial transcriptomics (ST), utilizing a gene panel designed to capture alTue, 20 Oct 2026 00:00:00 GMThttps://doi.org/10.1038/s41467-026-72139-810.1038/s41467-026-72139-8Nature Communications — Understanding the genetic regulation of RNA abundance is essential to defining disease mechanisms. However, conventional expression quantitative loci (eQTL) studies quantify RNA molecules across the transcript lifecycle. While most eQTL likely affect transcription by altering promoter or enhancer function within the nucleus, it is also possible that they modulate any processes after transcription, including chemical modifications and RNA stability in the cytosol. To elucidate distinct eQTL mechaTue, 20 Oct 2026 00:00:00 GMT